CN104468471B

CN104468471B - A kind of method and apparatus for being used to be grouped acoustic echo elimination

Info

Publication number: CN104468471B
Application number: CN201310419143.1A
Authority: CN
Inventors: 李舟洲; 蔡亦钢
Original assignee: Alcatel Optical Networks Israel Ltd
Current assignee: Alcatel Optical Networks Israel Ltd
Priority date: 2013-09-13
Filing date: 2013-09-13
Publication date: 2017-11-03
Anticipated expiration: 2033-09-13
Also published as: CN104468471A; WO2015036857A1

Abstract

It is an object of the invention to provide a kind of method and apparatus for being used to be grouped acoustic echo elimination.Echo cancellation devices obtain the source stream of voice packets of pending PAEC call ends；According to source stream of voice packets, the targeted packets stream of correspondence call ends in destination buffer is updated；According to the reference packet stream of the correspondence call ends in correspondence reference buffer area, combining target stream of packets and the direction of transfer information corresponding to each packet data package in reference packet stream, PAEC is carried out to the targeted packets stream, echo stream of packets and the corresponding end sent it in call ends have been eliminated to obtain.Compared with prior art, the present invention realizes two-way packet acoustic echo and eliminated, and exponentially improves the performance of PAEC channels, reduce hardware quantity and corresponding maintenance cost, call treatment and related signaling expense are decreased simultaneously, enters and supports that there is provided transparent PAEC functions without any signaling.

Description

A kind of method and apparatus for being used to be grouped acoustic echo elimination

Technical field

The present invention relates to the communications field, more particularly to a kind of technology for being used to be grouped acoustic echo elimination.

Background technology

Acoustic echo in mobile network is due to that the design of mobile phone or other hand free devices is not good, phonetic incepting Fang Yang The sound that sound device is sent is sent to recipient's microphone（And then send voice sender back to）Caused by.Acoustic echo is eliminated （Acoustic Echo Cancellation, AEC）The echo in signal of communication can be removed.It is to ensure logical that acoustic echo, which is eliminated, The core capabilities of sound quality in letter.

In a circuit switched network, traditional AEC technologies are removed in waveform domain to acoustic echo has been made fine. However, in packet network（Such as voice on IP network, VoIP）, also without the approved mode for being used to perform AEC.One A little suppliers（Such as Broadcom（With reference to US7333447）, Samsung, 3Com etc.）The AEC for packet network has been invented, but It is that this kind of AEC needs that stream of packets is first decoded into analog or digital signal（That is, it is transformed into waveform domain）, eliminated using conventional art Echo in signal, then recompiles back the signal for eliminating echo in packet（That is, packet domain is converted back）.Due to multiple Coding/decoding, result in sound quality（Voice quality, VQ）Decline so that counteract code conversion exempt operation （Transcoder Free Operation, TrFO）The advantage obtained on repeatedly coding and decoding is excluded.Further, since meter Complexity and huge buffer requirement are calculated, traditional AEC only supports limited tail length to postpone, therefore, traditional AEC is used for VoIP Efficiency is very low during network.

Alcatel-Lucent/AT＆T Labs（Alcatel-Lucent/Bell Labs）A kind of real point is invented Group domain acoustic echo is eliminated（Packet Acoustic Echo Cancellation, PAEC）Technology,（For example）Only need to use The parameter of waveform described in EVRC or EVRC-B packets, it becomes possible to which detection suppresses the acoustic echo in stream of packets.AT＆T Labs There are 3 related patents or patent application in PAEC fields：

-US7852792Packet Based Echo Cancellation and Suppression(granted on12/14/2010)by Binshi Cao et al.

-US008144862method and Apparatus for the Detection and Suppression of Echo in Packet based Communication Networks Using Frame Energy Estimation (granted on3/27/2012)by Binshi Cao et al.

-US2009/0168673Method and Apparatus for Detecting and Suppressing Echo in Packet Networks(published on7/2/2009)by Lampros Kalampoukas and Semyon Sosin.

In above-mentioned patent or patent application, it is compared and predicts by using the waveform characterising parameter of packet, will join Examine flow point group and be grouped in target stream in PAEC channels and compared, so as to remove（In target stream）Similar packet（It is identified as Echo）, realize the basic skills that elimination/suppression in packet network is grouped acoustic echo.

However, the method provided in these patents or patent application is only for unidirectional PAEC, and it can not provide two-way PAEC.One audio call is related to two or more correspondents, and the echo that eliminate each correspondent generation is more it is necessary to dispose Multiple unidirectional PAEC channels in individual unidirectional PAEC equipment or single PAEC equipment.From packet-switching performance and capacity point Analysis, especially in packet switching network inner exchanging scene, unidirectional PAEC limited capacity, and industrial quality may not reached With performance standard.PAEC products with unidirectional packet echo cancellor may not meet user's need in packet switch well Ask.Therefore, for the deployment of actual industry, these unidirectional PAEC methods all have the disadvantage in that and limited.

For example, Fig. 1, which illustrates a kind of unidirectional packet acoustic echo being described in US2009/0168673, eliminates knot Structure.One unidirectional PAEC channel can only distribute to a correspondent, and the direction that it needs to distinguish voice flow is " going to " or " come From " correspondent.If going to the correspondent, the voice flow is a reference stream.If from the correspondent, the voice flow It is a target stream.Or the voice flow is handled as reference packet and run, run or being handled as targeted packets.It is crucial It is that reference stream process part is run parallel when different with target stream process part.

The obvious of this unidirectional packet acoustic echo removing method has the disadvantage that efficiency low cost is high.Divide although possessing and referring to Group processing module can buffer the voice flow of other direction, but it does not run parallel with targeted packets processing module, does not also disappear Except the echo in reference packet.Two-way packet echo cancellor is realized, the signaling that two PAEC channels are provided and doubled is still needed to With management service expense.In a packet switching network inner exchanging scene, this is undoubtedly the waste to resource.

The content of the invention

It is an object of the invention to provide a kind of method and apparatus for being used to be grouped acoustic echo elimination.

According to an aspect of the invention, there is provided a kind of method for being used to be grouped acoustic echo elimination, wherein, this method Comprise the following steps：

A obtains the source stream of voice packets for the call ends that pending packet acoustic echo is eliminated, wherein, the source voice point Group stream includes one or more packets packet；

B updates the targeted packets stream of the correspondence call ends in destination buffer according to the source stream of voice packets, its In, the targeted packets stream includes the direction of transfer information corresponding to each packet data package in the targeted packets stream；

C is according to the reference packet stream for corresponding to the correspondence call ends in reference buffer area, with reference to the targeted packets stream With the direction of transfer information corresponding to each packet data package in the reference packet stream, the targeted packets stream is returned Sound is eliminated, to obtain the elimination echo stream of packets corresponding with the targeted packets stream；

D according to it is described eliminated echo stream of packets corresponding to direction of transfer information, eliminated echo stream of packets by described Send the corresponding end into the call ends.

According to another aspect of the present invention, a kind of echo cancellation devices for being used to be grouped acoustic echo elimination are additionally provided, Wherein, the equipment includes：

Acquisition device, the source stream of voice packets for obtaining the call ends that pending packet acoustic echo is eliminated, wherein, The source stream of voice packets includes one or more packets packet；

Target update device, for according to the source stream of voice packets, updating the call two of correspondence in destination buffer The targeted packets stream at end, wherein, the targeted packets stream is included corresponding to each packet data package in the targeted packets stream Direction of transfer information；

Cancellation element, for the reference packet stream according to the correspondence call ends in correspondence reference buffer area, with reference to institute Targeted packets stream and the direction of transfer information corresponding to each packet data package in the reference packet stream are stated, to the target Stream of packets carries out echo cancellor, to obtain the elimination echo stream of packets corresponding with the targeted packets stream；

Dispensing device, for having eliminated the direction of transfer information corresponding to echo stream of packets according to, has disappeared described Except echo stream of packets sends the corresponding end into the call ends.

Compared with prior art, the present invention is eliminated by obtaining pending packet acoustic echo in echo cancellation devices The source stream of voice packets of call ends, according to the source stream of voice packets, updates the correspondence call ends in destination buffer Targeted packets stream, according to the reference packet stream of the correspondence call ends in correspondence reference buffer area, with reference to the target point The targeted packets are flowed into by group stream and the direction of transfer information corresponding to each packet data package in the reference packet stream Row echo cancellor, to obtain the elimination echo stream of packets corresponding with the targeted packets stream, has been eliminated finally according to described Direction of transfer information corresponding to echo stream of packets, pair into the call ends is sent by the echo stream of packets that eliminated Ying Duan；It is achieved thereby that two-way packet acoustic echo is eliminated, exponentially improve the performance of PAEC channels, reduce hardware quantity with And corresponding maintenance cost, while decreasing call treatment and related signaling expense, enter and support that there is provided saturating without any signaling Bright PAEC functions.

Brief description of the drawings

By reading the detailed description made to non-limiting example made with reference to the following drawings, of the invention is other Feature, objects and advantages will become more apparent upon：

Fig. 1 shows a kind of unidirectional packet sound being described in US2009/0168673 according to one aspect of the invention Learn echo canceling structure schematic diagram；

Fig. 2 shows a kind of echo cancellation devices signal for being used to be grouped acoustic echo elimination according to one aspect of the invention Figure；

Fig. 3 shows that a kind of echo cancellor for being used to be grouped acoustic echo elimination in accordance with a preferred embodiment of the present invention is set Standby schematic diagram；

Fig. 4 shows a kind of method flow diagram for being used to be grouped acoustic echo elimination according to a further aspect of the present invention；

Fig. 5 shows a kind of method flow for being used to be grouped acoustic echo elimination in accordance with a preferred embodiment of the present invention Figure；

Fig. 6 shows that a kind of two-way packet acoustic echo according to a preferred embodiment of the present invention is eliminated with reference to signal Figure, wherein, the packet data package in each direction as another direction packet data package reference；

Fig. 7 shows that a kind of two-way packet acoustic echo according to a preferred embodiment of the present invention is eliminated with reference to signal Figure, wherein, the packet data package for eliminating echo in each direction as the packet data package in another direction reference；

Fig. 8 shows that one kind according to a preferred embodiment of the present invention is used as ginseng by the use of non-echo cancellor packet data package Buffering and comparison schematic diagram that the two-way packet acoustic echo examined is eliminated；

Fig. 9 shows that one kind according to a preferred embodiment of the present invention is used as reference by the use of echo cancellor packet data package Two-way packet acoustic echo eliminate buffering and comparison schematic diagram；

Figure 10 shows that a kind of comparison of echo frame for A ends according to a preferred embodiment of the present invention is calculated with removing Method；

Figure 11 shows that a kind of comparison of echo frame for B ends according to a preferred embodiment of the present invention is calculated with removing Method.

Same or analogous reference represents same or analogous part in accompanying drawing.

Embodiment

The present invention is described in further detail below in conjunction with the accompanying drawings.

Fig. 2 shows a kind of echo cancellation devices signal for being used to be grouped acoustic echo elimination according to one aspect of the invention Figure；Wherein, the echo cancellation devices include acquisition device 1, target update device 2, cancellation element 3, dispensing device 4.Specifically Ground, acquisition device 1 obtains the source stream of voice packets for the call ends that pending packet acoustic echo is eliminated, wherein, the source language Sound stream of packets includes one or more packets packet；Target update device 2 delays according to the source stream of voice packets, more fresh target The targeted packets stream of the correspondence call ends in area is rushed, wherein, the targeted packets stream is included in the targeted packets stream Direction of transfer information corresponding to each packet data package；Cancellation element 3 is according to the correspondence call in correspondence reference buffer area The reference packet stream at two ends, with reference to corresponding to each packet data package in the targeted packets stream and the reference packet stream Direction of transfer information, echo cancellor is carried out to the targeted packets stream, to obtain and the targeted packets stream is corresponding has disappeared Except echo stream of packets；Dispensing device 4 according to it is described eliminated echo stream of packets corresponding to direction of transfer information, disappeared described Except echo stream of packets sends the corresponding end into the call ends.

Here, the echo cancellation devices include but is not limited to according to the instruction for being previously set or storing can automatically enter Row numerical computations and the electronic hardware or software equipment of information processing；Wherein, the hardware device is including but not limited to micro- Processor, application specific integrated circuit (ASIC), programmable gate array（FPGA）, digital processing unit（DSP）, embedded device etc..This Art personnel will be understood that other echo cancellation devices are equally applicable to the present invention, should also be included in present invention protection Within scope, and it is incorporated herein by reference herein.

The echo cancellation devices can be used in any VOIP networks, Real Time Communication Network RTC and LTE/EPC In network, above-mentioned network is currently also without effective and generally acknowledged packet acoustic echo abatement apparatus.

Constantly worked between above-mentioned each device, here, it will be understood by those skilled in the art that " lasting " refers to State each device respectively in real time, or according to the mode of operation requirement of setting or real-time adjustment, carry out the source language of call ends Acquisition, the renewal of targeted packets stream, the acquisition for having eliminated echo stream of packets, the transmission for having eliminated echo stream of packets of sound stream of packets Deng until the echo cancellation devices stop obtaining the source packets of voice for the call ends that pending packet acoustic echo is eliminated Stream.

Acquisition device 1 obtains the source stream of voice packets for the call ends that pending packet acoustic echo is eliminated, wherein, it is described Source stream of voice packets includes one or more packets packet.

Specifically, the acquisition device 1 is from the call ends conversed（By taking communicating end A and communicating end B as an example）, obtain The source stream of voice packets for the call ends that pending packet acoustic echo is eliminated；Wherein, the source stream of voice packets is included from logical End A to communicating end B source stream of voice packets is talked about, also including the source stream of voice packets from communicating end B to communicating end A.Wherein, it is described One or more packets packet is included in the stream of voice packets of source（packet）, and the packet data package of the source stream of voice packets In may comprising echo bag.

Target update device 2 updates the correspondence call ends in destination buffer according to the source stream of voice packets Targeted packets stream, wherein, the targeted packets stream includes the biography corresponding to each packet data package in the targeted packets stream Send directional information.

Specifically, the target update device 2 passes through according to source stream of voice packets acquired in the acquisition device 1 Source stream of voice packets is sent to destination buffer, so that using the source stream of voice packets to the target in destination buffer point Group stream is updated, wherein, because source stream of voice packets is the voice point for the call ends that pending packet acoustic echo is eliminated Group stream, therefore, also includes the stream of voice packets corresponding to the call ends in the targeted packets stream.Here, the target Stream of packets includes the direction of transfer information corresponding to each packet data package in the targeted packets stream.

Preferably, the target update device 2 can determine the source stream of voice packets according to the source stream of voice packets In each packet data package corresponding to direction of transfer information；According to the source stream of voice packets, with reference to the source voice point The direction of transfer information of packet data package in group stream, updates the targeted packets of the correspondence call ends in destination buffer Stream.

Specifically, the target update device 2 can be according to the source stream of voice packets, by according to the source voice point Source address and destination address in group stream in the header packet information of each packet, calculating are determined corresponding to each packet data package Direction of transfer information.

For example, illustrate call ends by taking communicating end A and communicating end B as an example, then the direction of transfer information include from A to B or from B to A, if known communicating end A address and/or communicating end B address, according to the header packet information of the packet In source address and destination address, can directly determine the direction of transfer information corresponding to the packet；

Or, for example, by using predetermined calculating function, by the source address and mesh in the header packet information of the packet Address be compared, if source address be more than destination address, it is determined that the direction of transfer of the packet be from A to B, conversely, If source address is less than destination address, it is determined that the direction of transfer of the packet is from B to A, if in the presence of other situations, occurring Mistake, the packet is dropped.

The target update device 2 is according to the source stream of voice packets, with reference to the packet count in the source stream of voice packets According to the direction of transfer information of bag, the targeted packets stream of the correspondence call ends in destination buffer, therefore, the target are updated Stream of packets includes the targeted packets stream from A to B and the targeted packets stream from B to A.

Preferably, the target update device 2 can update correspondence in destination buffer according to the source stream of voice packets The targeted packets stream of the call ends；According to the targeted packets stream, each packet count in the targeted packets stream is determined According to the direction of transfer information corresponding to bag.

Specifically, the target update device 2 can be according to the source stream of voice packets, first to correspondence in destination buffer The targeted packets stream of the call ends is updated；Then further according to the targeted packets stream, by according to the target point Source address and destination address in group stream in the header packet information of each packet, calculating is determined every in the targeted packets stream Direction of transfer information corresponding to individual packet data package.Here, the computational methods with the target update device 2 according to described Source stream of voice packets, the method for determining the direction of transfer information corresponding to each packet data package in the source stream of voice packets It is same or similar, therefore will not be repeated here, and be incorporated herein by reference.

Cancellation element 3 is according to the reference packet stream for corresponding to the correspondence call ends in reference buffer area, with reference to the mesh Stream of packets and the direction of transfer information corresponding to each packet data package in the reference packet stream are marked, to the targeted packets Stream carries out echo cancellor, to obtain the elimination echo stream of packets corresponding with the targeted packets stream.

Specifically, the cancellation element 3 obtains corresponding described logical in reference buffer area corresponding with the destination buffer The reference packet stream at two ends is talked about, wherein, the voice that the reference packet stream can be wrapped according to source stream of voice packets with echo Stream of packets determined, or, it can be carried out not including echo after packet acoustic echo is eliminated according to the source stream of voice packets The stream of voice packets of bag is determined；The cancellation element 3 according to the targeted packets stream with it is each in the reference packet stream Direction of transfer information corresponding to packet data package, the targeted packets stream of different directions and the reference packet stream are carried out Contrast, for example, will be contrasted from the targeted packets stream at A ends to B ends with the reference packet stream from B ends to A ends, or will be from B The targeted packets stream to A ends is held to be contrasted with the reference packet stream from A ends to B ends, based on packet acoustic echo cancellation algorithm （PAEC algorithms）To detect whether comprising echo bag in the targeted packets stream, if comprising echo bag, by deleting described return Sound bag is disappeared using replacing bag and the mode such as being replaced to detected echo bag and carry out echo to the targeted packets stream Remove.Specifically, for example, being replaced using bag is replaced to detected echo bag, to obtain and the targeted packets stream phase It is corresponding to have eliminated echo stream of packets.Wherein, the bag of replacing includes but is not limited to noise bag（For example, including certain type The packet of noise, such as white noise, comfort noise）, noiseless bag（For example, space division group）, finally cache in targeted packets stream 1/8th rate packets etc., and its mixing.

Here, the determination method of the direction of transfer information corresponding to each packet data package in the reference packet stream, With determining that the method for the direction of transfer information corresponding to each packet data package in the source stream of voice packets is same or similar, Therefore will not be repeated here, and be incorporated herein by reference.

Dispensing device 4 according to it is described eliminated echo stream of packets corresponding to direction of transfer information, eliminated back described Sound stream of packets sends the corresponding end into the call ends.

Specifically, the dispensing device 4 according to it is described eliminated echo stream of packets corresponding to direction of transfer information, for example According to the destination address information for having eliminated echo stream of packets, or according to call corresponding in the direction of transfer information Client information, by it is described eliminated echo stream of packets send to it is corresponding corresponding to the source for having eliminated echo stream of packets End.

If for example, the direction of transfer information eliminated corresponding to echo stream of packets is A ends to B ends, by described in Eliminate echo stream of packets to send to B ends, here, B ends are the corresponding end at A ends.

So as to which the present invention realizes a kind of two-way packet acoustic echo removing method, this method：

- reduce hardware quantity and corresponding maintenance cost：Compared with unidirectional PAEC, two-way PAEC hsrdware requirements halve and saved About related maintenance；

- reduce call treatment and signaling consumption：Only need to distribute a PAEC channel for basic call；

- realize implicit/transparent PAEC supported without any signaling：In packet voice（Transmission）Gateway energy in path Two-way PAEC is enough integrated, to provide implicitly/transparent PAEC for side a and b.

Preferably, the source stream of voice packets can be respectively sent to the Target buffer by the target update device 2 Area and reference buffer area, to update the targeted packets stream of the correspondence call ends in the mark buffering area, and the reference The reference packet stream of the correspondence call ends in buffering area, wherein, the targeted packets stream is included in the targeted packets stream Each packet data package corresponding to direction of transfer information, the reference packet stream include the reference packet stream in it is each Direction of transfer information corresponding to packet data package.

Specifically, the target update device 2 is according to source stream of voice packets acquired in the acquisition device 1, by institute The source stream of voice packets of stating is respectively sent to the destination buffer and reference buffer area, using the source stream of voice packets, to institute State the targeted packets stream in destination buffer and the reference packet stream in reference buffer area is updated；Wherein, due to source language Sound stream of packets be it is pending packet acoustic echo eliminate call ends stream of voice packets, therefore, the targeted packets stream with Include the stream of voice packets corresponding to the call ends in reference packet stream.Here, the targeted packets stream is comprising described The direction of transfer information corresponding to each packet data package in targeted packets stream, the reference packet stream includes the reference point The direction of transfer information corresponding to each packet data package in group stream.

Here, the targeted packets stream and the determination side of the direction of transfer information of the packet data package in reference packet stream Method, or phase identical with the method for determining the direction of transfer information corresponding to each packet data package in the source stream of voice packets Seemingly, thus will not be repeated here, and be incorporated herein by reference.

For example, Fig. 6 shows that a kind of two-way packet acoustic echo according to a preferred embodiment of the present invention eliminates reference Schematic diagram, wherein, the packet data package in each direction as another direction packet data package reference.

Specifically, the source stream of voice packets come from A ends and/or B ends is all sent to reference packet processing by RTP resolvers And in targeted packets processing, in the buffering area of separation（Destination buffer and reference buffer area）Middle buffering targeted packets stream and ginseng Examine stream of packets.Here, the source stream of voice packets transmitted by RTP resolvers includes the packet data package of the source stream of voice packets Load and head.Wherein, in the source stream of voice packets sent from A ends or with B ends echo or not comprising echo, from B Hold in the source stream of voice packets sent or with A ends echo or not comprising echo.Because the targeted packets stream is logical Cross and cache the source stream of voice packets and determined, therefore, if including echo, the targeted packets in the source stream of voice packets Also corresponding echo is included in stream；Do not include in echo, the targeted packets stream if not including in the source packets of voice yet Corresponding echo.

In the destination buffer and the reference buffer area, the targeted packets stream is included in the targeted packets stream Each packet data package corresponding to direction of transfer information, the reference packet stream include the reference packet stream in it is each Direction of transfer information corresponding to packet data package.

In the PAEC algoritic modules, the targeted packets stream in a direction in the destination buffer, with the reference The reference packet stream of the other direction prestored in buffering area is contrasted, as shown in figure 8, packet collection in targeted packets stream Close（Packet j to packet j+M, i.e. B ends to A extreme directions targeted packets stream）Set corresponding with reference packet stream respectively 1st, set 2 ..., set K（That is the stream of voice packets at A ends to B ends, for carrying out B ends to the reference of A extreme directions）Progress pair Than packet set in targeted packets stream（Packet i to packet i+N, i.e. A ends to B extreme directions targeted packets stream）Respectively It is corresponding with reference packet stream set 1, set 2 ..., set Q（That is the stream of voice packets at B ends to A ends, for carrying out A ends To the reference of B extreme directions）Contrasted, whether there is echo bag in the targeted packets stream to determine different directions.Wherein, Wrapped in the reference packet stream comprising corresponding echo.

If there is echo bag in the targeted packets stream, the PAEC algoritic modules carry out packet acoustic echo to it and disappeared Except calculating, the echo stream of packets of elimination eliminated after echo is respectively sent to A ends and B ends.

Preferably, the echo cancellation devices also include referring to updating device（It is not shown）, wherein, it is described to refer to more new clothes Putting can be according to the reference packet stream for having eliminated echo stream of packets, having updated in the reference buffer area.

Specifically, the reference updating device can be interacted with the cancellation element 3, and described echo has been eliminated to obtain Stream of packets；Then, the reference updating device has eliminated echo stream of packets according to described, to the reference in the reference buffer area Stream of packets is updated；It is used as the ginseng being compared with the targeted packets stream so as to eliminate echo stream of packets described in Stream of packets is examined, the use to buffering area is reduced, with preferably effect is referred to, so as to further increase the accurate of PAEC Rate.

For example, Fig. 7 shows that a kind of two-way packet acoustic echo according to a preferred embodiment of the present invention eliminates reference Schematic diagram, wherein, the packet data package for eliminating echo in each direction as the packet data package in another direction ginseng Examine.

Specifically, the source stream of voice packets come from A ends and/or B ends is all sent to targeted packets processing by RTP resolvers In, the targeted packets stream is determined by caching the source stream of voice packets, therefore, is also included in the targeted packets stream Corresponding echo.Here, the source stream of voice packets transmitted by RTP resolvers includes the grouped data of the source stream of voice packets The load of bag and head.Wherein, in the source stream of voice packets sent from A ends or with B ends echo or not comprising echo, In the source stream of voice packets sent from B ends or with A ends echo or not comprising echo.

Reference packet processing interacts with PAEC algoritic modules, has been eliminated with obtaining determined by the PAEC algoritic modules Echo stream of packets, and the echo stream of packets that eliminated is buffered to the reference buffer area, to be used as the reference packet stream.

Here, each packet data package in the targeted packets stream and the reference packet stream is comprising corresponding to it Direction of transfer information.

In the PAEC algoritic modules, the targeted packets stream in a direction in the destination buffer, with the reference The reference packet stream of the other direction prestored in buffering area is contrasted, as shown in figure 9, packet collection in targeted packets stream Close（Packet j to packet j+M, i.e. B ends to A extreme directions targeted packets stream）Set corresponding with reference packet stream respectively 1st, set 2 ..., set K（That is the stream of voice packets at A ends to B ends, for carrying out B ends to the reference of A extreme directions）Progress pair Than packet set in targeted packets stream（Packet i to packet i+N, i.e. A ends to B extreme directions targeted packets stream）Respectively It is corresponding with reference packet stream set 1, set 2 ..., set Q（That is the stream of voice packets at B ends to A ends, for carrying out A ends To the reference of B extreme directions）Contrasted, whether there is echo bag in the targeted packets stream to determine different directions.Wherein, No longer wrapped in the reference packet stream comprising corresponding echo, belong to and eliminated echo stream of packets.

Here, respectively illustrating a kind of comparison for A ends and the echo frame at B ends with reference to Fig. 8 or Fig. 9, Figure 10 and Figure 11 With removing algorithm.

Specifically, in Fig. 10, " N+1 " is the target window size for direction A to B, and " N+Q " is corresponding reference window Mouth size." Q " according to the echo path delay at B ends by being determined.Represent the N+1 in the destination buffer from A to B（i, i+1,…,i+N）N+1 in individual packet and the reference buffer area from A to B（q,q+1,…,q+N）The contrast knot of individual packet Really.Here, those skilled in the art will be understood that the sender of the reference packet stream for as targeted packets stream A to B It is should be to information from B to A.The minimum value of (q=q, q+1 ..., q+Q-1) will be with minimum threshold e_THCompare, to determine With the presence or absence of echo.

（Formula 1）

（Formula 2）

Result minimum value in formula 2 represents the similitude of direction A to B targets stream and corresponding reference stream；If formula 2 As a result following formula is met：

（Formula 3）

Then illustrate that targeted packets stream includes echo with existing in reference packet stream in similitude, therefore targeted packets stream.

In fig. 11, " M+1 " is the target window size for direction B to A, and " M+K " is corresponding reference windows size. " K " according to the echo path delay at A ends by being determined.Represent the M+1 in the destination buffer from B to A（j,j+1,…, j+M）M+1 in individual packet and the reference buffer area from B to A（k,k+1,…,k+M）The comparing result of individual packet.Here, Those skilled in the art will be understood that the direction of transfer information of the reference packet stream for as targeted packets stream B to A should For from A to B.The minimum value of (k=k, k+1 ..., k+Q-1) will be with minimum threshold e_THCompare, to determine whether there is Echo.

（Formula 4）

（Formula 5）

Result minimum value in formula 5 represents the similitude of direction B to A targets stream and corresponding reference stream；If formula 5 As a result following formula is met：

（Formula 6）

Here, P is LSP value（Line Spectral Pair, line spectrum pair）.

Fig. 3 shows that a kind of echo cancellor for being used to be grouped acoustic echo elimination in accordance with a preferred embodiment of the present invention is set Standby schematic diagram；Wherein, the echo cancellation devices include acquisition device 1 ', target update device 2 ', cancellation element 3 ', transmission dress 4 ' are put, wherein, the cancellation element 3 ' includes echo determining unit 31 ', echo cancellation unit 32 '.Specifically, acquisition device 1 ' The source stream of voice packets for the call ends that pending packet acoustic echo is eliminated is obtained, wherein, the source stream of voice packets is included One or more packets packet；Target update device 2 ' updates correspondence in destination buffer according to the source stream of voice packets The targeted packets stream of the call ends, wherein, the targeted packets stream includes each packet count in the targeted packets stream According to the direction of transfer information corresponding to bag；Echo determining unit 31 ' is according to the correspondence call ends in correspondence reference buffer area Reference packet stream, the transmission with reference to corresponding to each packet data package in the targeted packets stream and the reference packet stream Whether directional information, determine in the targeted packets stream comprising echo bag；Echo cancellation unit 32 ' is when in the targeted packets stream Comprising echo bag, echo cancellor is carried out to the targeted packets stream, to obtain the elimination corresponding with the targeted packets stream Echo stream of packets；Dispensing device 4 ' has eliminated the direction of transfer information corresponding to echo stream of packets according to, has disappeared described Except echo stream of packets sends the corresponding end into the call ends.

Wherein, shown in the acquisition device 1 ' of the echo cancellation devices, target update device 2 ', dispensing device 4 ' and Fig. 2 Corresponding intrument is identical or essentially identical, therefore here is omitted, and is incorporated herein by reference.

Constantly worked between above-mentioned each device, here, it will be understood by those skilled in the art that " lasting " refers to State each device respectively in real time, or according to the mode of operation requirement of setting or real-time adjustment, carry out the source language of call ends The acquisition of sound stream of packets, the renewal of targeted packets stream, the determination whether wrapped comprising echo, the acquisition for having eliminated echo stream of packets, Transmission of echo stream of packets etc. is eliminated, until the echo cancellation devices stop obtaining what pending packet acoustic echo was eliminated The source stream of voice packets of call ends.

Echo determining unit 31 ' corresponds to the reference packet stream of the call ends according to corresponding in reference buffer area, with reference to The targeted packets stream and the direction of transfer information corresponding to each packet data package in the reference packet stream, it is determined that described Whether echo bag is included in targeted packets stream.

Specifically, the echo determining unit 31 ' obtains corresponding in reference buffer area corresponding with the destination buffer The reference packet stream of the call ends, wherein, the reference packet stream can be according to source stream of voice packets with echo bag Stream of voice packets determined, or, can according to the source stream of voice packets carry out packet acoustic echo eliminate after not wrapping The stream of voice packets wrapped containing echo is determined；The echo determining unit 31 ' is according to the targeted packets stream and the reference point The direction of transfer information corresponding to each packet data package in group stream, by the targeted packets stream of different directions and the ginseng Examination mark group stream is contrasted, for example, by from the targeted packets stream at A ends to B ends with being carried out pair from the reference packet stream at B ends to A ends Than, or will be contrasted from the targeted packets stream at B ends to A ends with the reference packet stream from A ends to B ends, based on packet acoustics Echo cancellation algorithm（PAEC algorithms）Whether to detect in the targeted packets stream comprising echo bag.

Preferably, the echo determining unit 31 ' is according to the reference for corresponding to the correspondence call ends in reference buffer area Stream of packets, the direction of transfer with reference to corresponding to each packet data package in the targeted packets stream and the reference packet stream is believed Breath, and it is corresponding with multiple consecutive packet data packages corresponding in the reference packet stream with the targeted packets stream Energy hierarchical information, whether determine in the targeted packets stream comprising echo bag.

Specifically, according to Figure 10, Figure 11 and formula 1 to formula 6, LSP is based on the invention provides one kind（Line Spectral Pair, line spectrum pair）The method for carrying out the detection of echoes of packet voice；If the echo determining unit 31 ' is according to right The reference packet stream of the correspondence call ends in reference buffer area is answered, with reference to the targeted packets stream and the reference packet stream In each packet data package corresponding to direction of transfer information, using this method, determine the targeted packets stream with reference to point Group stream in have similar packet, then the echo determining unit 31 ' can further combined with the targeted packets stream with it is described The corresponding energy hierarchical information of corresponding multiple consecutive packet data packages in reference packet stream（I.e. all kinds of gains （gain）Information）, the similar packet in the targeted packets stream is judged with the presence or absence of decay, i.e. energy hierarchical information is low In the energy hierarchical information of the corresponding reference packet stream；If in the presence of it is echo bag to prove the similar packet, then Wrapped in the targeted packets stream comprising echo.

This is due to that in echo, echo energy typically has a certain degree of decay than original words sound, so that by energy layer Level compares the subsidiary conditions as detection echo bag.

Echo cancellation unit 32 ' carries out echo when being wrapped in the targeted packets stream comprising echo to the targeted packets stream Eliminate, to obtain the elimination echo stream of packets corresponding with the targeted packets stream.

Specifically, if being wrapped in the targeted packets stream comprising echo, echo cancellation unit 32 ' is then to the targeted packets stream Echo cancellor is carried out, for example, being replaced using bag is replaced to detected echo bag, to obtain and the targeted packets stream The corresponding echo stream of packets of elimination.Wherein, the bag of replacing includes but is not limited to noise bag（For example, including certain type Noise packet, such as white noise, comfort noise）, noiseless bag（For example, space division group）, finally delay in targeted packets stream / 8th rate packets deposited etc., and its mixing.

Here, when utilization carries the replacement bag to fixed load, it is necessary to correspondingly change RTP and other length related words Section and verification, for example, the specific head of modification platform, IP, UDP, RTP heads.

Fig. 4 shows a kind of method flow diagram for being used to be grouped acoustic echo elimination according to a further aspect of the present invention.Tool Body, in step s1, echo cancellation devices obtain the source packets of voice for the call ends that pending packet acoustic echo is eliminated Stream, wherein, the source stream of voice packets includes one or more packets packet；In step s2, echo cancellation devices according to The source stream of voice packets, updates the targeted packets stream of the correspondence call ends in destination buffer, wherein, the target point Group stream includes the direction of transfer information corresponding to each packet data package in the targeted packets stream；In step s3, echo Abatement apparatus according to the reference packet stream of the correspondence call ends in correspondence reference buffer area, with reference to the targeted packets stream with The direction of transfer information corresponding to each packet data package in the reference packet stream, echo is carried out to the targeted packets stream Eliminate, to obtain the elimination echo stream of packets corresponding with the targeted packets stream；In step s4, echo cancellation devices root According to the direction of transfer information eliminated corresponding to echo stream of packets, the echo stream of packets that eliminated is sent to described logical Talk about the corresponding end in two ends.

Constantly worked between above steps, here, it will be understood by those skilled in the art that " lasting " refers to State each step respectively in real time, or according to the mode of operation requirement of setting or real-time adjustment, carry out the source language of call ends Acquisition, the renewal of targeted packets stream, the acquisition for having eliminated echo stream of packets, the transmission for having eliminated echo stream of packets of sound stream of packets Deng until the echo cancellation devices stop obtaining the source packets of voice for the call ends that pending packet acoustic echo is eliminated Stream.

In step s1, echo cancellation devices obtain the source voice point for the call ends that pending packet acoustic echo is eliminated Group stream, wherein, the source stream of voice packets includes one or more packets packet.

Specifically, in step s1, echo cancellation devices are from the call ends conversed（With communicating end A and communicating end B Exemplified by）, obtain the source stream of voice packets for the call ends that pending packet acoustic echo is eliminated；Wherein, the source packets of voice Stream includes the source stream of voice packets from communicating end A to communicating end B, also including the source packets of voice from communicating end B to communicating end A Stream.Wherein, one or more packets packet is included in the source stream of voice packets（packet）, and the source stream of voice packets Packet data package in may comprising echo bag.

In step s2, echo cancellation devices update correspondence in destination buffer described according to the source stream of voice packets The targeted packets stream of call ends, wherein, the targeted packets stream includes each packet data package in the targeted packets stream Corresponding direction of transfer information.

Specifically, in step s2, echo cancellation devices lead to according to source stream of voice packets acquired in the step s1 Cross and send source stream of voice packets to destination buffer, so that using the source stream of voice packets to the target in destination buffer Stream of packets is updated, wherein, because source stream of voice packets is the voice for the call ends that pending packet acoustic echo is eliminated Stream of packets, therefore, also includes the stream of voice packets corresponding to the call ends in the targeted packets stream.Here, the mesh Mark stream of packets and include the direction of transfer information corresponding to each packet data package in the targeted packets stream.

Preferably, in step s2, echo cancellation devices can determine the source voice according to the source stream of voice packets The direction of transfer information corresponding to each packet data package in stream of packets；According to the source stream of voice packets, with reference to the source The direction of transfer information of packet data package in stream of voice packets, updates the target of the correspondence call ends in destination buffer Stream of packets.

Specifically, in step s2, echo cancellation devices can be according to the source stream of voice packets, by according to the source Each packet data package is determined in source address and destination address in the header packet information of the packet of each in stream of voice packets, calculating Corresponding direction of transfer information.

In step s2, echo cancellation devices are according to the source stream of voice packets, with reference in the source stream of voice packets The direction of transfer information of packet data package, updates the targeted packets stream of the correspondence call ends in destination buffer, therefore, institute Stating targeted packets stream includes the targeted packets stream from A to B and the targeted packets stream from B to A.

Preferably, in step s2, echo cancellation devices can update destination buffer according to the source stream of voice packets The targeted packets stream of the middle correspondence call ends；According to the targeted packets stream, determine each in the targeted packets stream Direction of transfer information corresponding to packet data package.

Specifically, in step s2, echo cancellation devices can be according to the source stream of voice packets, first to destination buffer The targeted packets stream of the middle correspondence call ends is updated；Then further according to the targeted packets stream, by according to described The targeted packets stream is determined in source address and destination address in targeted packets stream in the header packet information of each packet, calculating In each packet data package corresponding to direction of transfer information.Here, according to institute in the computational methods and the step s2 Source stream of voice packets is stated, the side of the direction of transfer information corresponding to each packet data package in the source stream of voice packets is determined Method is same or similar, therefore will not be repeated here, and is incorporated herein by reference.

In step s3, echo cancellation devices are according to the reference packet for corresponding to the correspondence call ends in reference buffer area Stream, the direction of transfer information with reference to corresponding to each packet data package in the targeted packets stream and the reference packet stream, Echo cancellor is carried out to the targeted packets stream, to obtain the elimination echo stream of packets corresponding with the targeted packets stream.

Specifically, in step s3, echo cancellation devices are obtained in reference buffer area corresponding with the destination buffer The reference packet stream of the correspondence call ends, wherein, the reference packet stream can carrying back according to source stream of voice packets The stream of voice packets of sound bag determined, or, it can be carried out according to the source stream of voice packets after packet acoustic echo eliminates The stream of voice packets wrapped not comprising echo is determined；In step s3, echo cancellation devices are according to the targeted packets stream and institute The direction of transfer information corresponding to each packet data package in reference packet stream is stated, by the targeted packets stream of different directions Contrasted with the reference packet stream, for example, by from the targeted packets stream at A ends to B ends and reference packet from B ends to A ends Stream is contrasted, or will be contrasted from the targeted packets stream at B ends to A ends with the reference packet stream from A ends to B ends, is based on It is grouped acoustic echo cancellation algorithm（PAEC algorithms）To detect whether comprising echo bag in the targeted packets stream, if including echo Bag, then by deleting echo bag or the mode such as being replaced to detected echo bag to the mesh using bag is replaced Mark stream of packets and carry out echo cancellor.Specifically, for example, being replaced using bag is replaced to detected echo bag, to obtain The elimination echo stream of packets corresponding with the targeted packets stream.Wherein, the bag of replacing includes but is not limited to noise bag（Example Such as, the packet of the noise comprising certain type, such as white noise, comfort noise）, noiseless bag（For example, space division group）, in target / 8th rate packets finally cached in stream of packets etc., and its mixing.

In step s4, echo cancellation devices according to it is described eliminated echo stream of packets corresponding to direction of transfer information, The echo stream of packets that eliminated is sent into the corresponding end into the call ends.

Specifically, in step s4, echo cancellation devices according to it is described eliminated echo stream of packets corresponding to sender To information, for example, the destination address information of echo stream of packets is eliminated according to, or according in the direction of transfer information Corresponding call client information, by it is described eliminated echo stream of packets send to the source institute for having eliminated echo stream of packets Corresponding corresponding end.

Preferably, in step s2, the source stream of voice packets can be respectively sent to the mesh by echo cancellation devices Buffering area and reference buffer area are marked, to update the targeted packets stream of the correspondence call ends in the mark buffering area, Yi Jisuo The reference packet stream of the correspondence call ends in reference buffer area is stated, wherein, the targeted packets stream includes the target point The direction of transfer information corresponding to each packet data package in group stream, the reference packet stream is included in the reference packet stream Each packet data package corresponding to direction of transfer information.

Specifically, in step s2, echo cancellation devices, will according to source stream of voice packets acquired in the step s1 The source stream of voice packets is respectively sent to the destination buffer and reference buffer area, right using the source stream of voice packets The reference packet stream in targeted packets stream and reference buffer area in the destination buffer is updated；Wherein, due to source Stream of voice packets is the stream of voice packets for the call ends that pending packet acoustic echo is eliminated, therefore, the targeted packets stream With including the stream of voice packets corresponding to the call ends in reference packet stream.Here, the targeted packets stream includes institute The direction of transfer information corresponding to each packet data package in targeted packets stream is stated, the reference packet stream includes the reference The direction of transfer information corresponding to each packet data package in stream of packets.

Specifically, the source stream of voice packets come from A ends and/or B ends is sent to reference packet processing by RTP resolvers simultaneously And in targeted packets processing, in the buffering area of separation（Destination buffer and reference buffer area）Middle buffering targeted packets stream and ginseng Examine stream of packets.Here, the source stream of voice packets transmitted by RTP resolvers includes the packet data package of the source stream of voice packets Load and head.Wherein, in the source stream of voice packets sent from A ends or with B ends echo or not comprising echo, from B Hold in the source stream of voice packets sent or with A ends echo or not comprising echo.Because the targeted packets stream is logical Cross and cache the source stream of voice packets and determined, therefore, if including echo, the targeted packets in the source stream of voice packets Also corresponding echo is included in stream；Do not include in echo, the targeted packets stream if not including in the source packets of voice yet Corresponding echo.

Preferably, the echo cancellation devices also include step s5（It is not shown）, wherein, in step s5, echo cancellor Equipment can be according to the reference packet stream for having eliminated echo stream of packets, having updated in the reference buffer area.

Specifically, in step s5, echo cancellation devices can interact with the step s3, to obtain described eliminated Echo stream of packets；Then, in step s5, echo cancellation devices have eliminated echo stream of packets according to described, to described with reference to slow The reference packet stream rushed in area is updated；It is used as and the targeted packets stream so as to eliminate echo stream of packets described in The reference packet stream being compared, reduces the use to buffering area, with preferably effect is referred to, so as to further increase PAEC accuracy rate.

For example, Fig. 7 shows that a kind of two-way packet acoustic echo according to a preferred embodiment of the present invention eliminates reference Schematic diagram, wherein, the packet data package for eliminating echo in each direction is used as another reflective packet count of direction band According to the reference of bag.

Specifically, the source stream of voice packets come from A ends and/or B ends is sent in targeted packets processing by RTP resolvers, The targeted packets stream is determined by caching the source stream of voice packets, therefore, and phase is also included in the targeted packets stream The echo answered.Here, the source stream of voice packets transmitted by RTP resolvers includes the packet data package of the source stream of voice packets Load and head.Wherein, in the source stream of voice packets sent from A ends or with B ends echo or not comprising echo, from B Hold in the source stream of voice packets sent or with A ends echo or not comprising echo.

Specifically, in Fig. 10, " N+1 " is the target window size for direction A to B, and " N+Q " is corresponding reference window Mouth size." Q " according to the echo path delay at B ends by being determined.Represent the N+1 in the destination buffer from A to B （i,i+1,…,i+N）N+1 in individual packet and the reference buffer area from A to B（q,q+1,…,q+N）The contrast of individual packet As a result.Here, those skilled in the art will be understood that the transmission of the reference packet stream for as targeted packets stream A to B Directional information should be from B to A.The minimum value of (q=q, q+1 ..., q+Q-1) will be with minimum threshold e_THCompare, with true Surely it whether there is echo.

（Formula 7）

（Formula 8）

Result minimum value in formula 8 represents the similitude of direction A to B targets stream and corresponding reference stream；If formula 8 As a result following formula is met：

（Formula 9）

（Formula 10）

（Formula 11）

Result minimum value in formula 11 represents the similitude of direction B to A targets stream and corresponding reference stream；If formula 11 Result meet following formula：

（Formula 12）

Here, P is LSP value（Line Spectral Pair, line spectrum pair）.

Fig. 5 shows a kind of method flow for being used to be grouped acoustic echo elimination in accordance with a preferred embodiment of the present invention Figure.Specifically, in step s1 ', echo cancellation devices obtain the source language for the call ends that pending packet acoustic echo is eliminated Sound stream of packets, wherein, the source stream of voice packets includes one or more packets packet；In step s2 ', echo cancellor is set It is standby that the targeted packets stream of the correspondence call ends in destination buffer is updated according to the source stream of voice packets, wherein, it is described Targeted packets stream includes the direction of transfer information corresponding to each packet data package in the targeted packets stream；In step s31 ' In, echo cancellation devices are according to the reference packet stream for corresponding to the correspondence call ends in reference buffer area, with reference to the target Stream of packets and the direction of transfer information corresponding to each packet data package in the reference packet stream, determine the targeted packets Whether echo bag is included in stream；In step s32 ', echo cancellation devices are worked as comprising echo bag in the targeted packets stream, to institute State targeted packets stream and carry out echo cancellor, to obtain the elimination echo stream of packets corresponding with the targeted packets stream；In step In rapid s4 ', echo cancellation devices eliminated echo stream of packets according to corresponding to direction of transfer information, eliminated described Echo stream of packets sends the corresponding end into the call ends.

Wherein, the step s1 ' of methods described, step s2 ', step s4 ' be identical with correspondence step shown in Fig. 4 or basic phase Together, thus here is omitted, and be incorporated herein by reference.

Constantly worked between above steps, here, it will be understood by those skilled in the art that " lasting " refers to State each step respectively in real time, or according to the mode of operation requirement of setting or real-time adjustment, carry out the source language of call ends The acquisition of sound stream of packets, the renewal of targeted packets stream, the determination whether wrapped comprising echo, the acquisition for having eliminated echo stream of packets, Transmission of echo stream of packets etc. is eliminated, until the echo cancellation devices stop obtaining what pending packet acoustic echo was eliminated The source stream of voice packets of call ends.

In step s31 ', echo cancellation devices divide according to the reference of the correspondence call ends in correspondence reference buffer area Group stream, the direction of transfer with reference to corresponding to each packet data package in the targeted packets stream and the reference packet stream is believed Whether breath, determine in the targeted packets stream comprising echo bag.

Specifically, in step s31 ', echo cancellation devices obtain reference buffer area corresponding with the destination buffer The reference packet stream of the middle correspondence call ends, wherein, the reference packet stream can carrying according to source stream of voice packets Echo bag stream of voice packets determined, or, can according to the source stream of voice packets carry out packet acoustic echo elimination after Do not include echo wrap stream of voice packets determined；In step s31 ', echo cancellation devices are according to the targeted packets stream With the direction of transfer information corresponding to each packet data package in the reference packet stream, the target of different directions is divided Group stream is contrasted with the reference packet stream, for example, by from the targeted packets stream at A ends to B ends and reference from B ends to A ends Stream of packets is contrasted, or will be contrasted from the targeted packets stream at B ends to A ends with the reference packet stream from A ends to B ends, Based on packet acoustic echo cancellation algorithm（PAEC algorithms）Whether to detect in the targeted packets stream comprising echo bag.

Preferably, in step s31 ', echo cancellation devices are according to the correspondence call ends in correspondence reference buffer area Reference packet stream, the transmission with reference to corresponding to each packet data package in the targeted packets stream and the reference packet stream Directional information, and with multiple consecutive packet data packages corresponding in the targeted packets stream and the reference packet stream Whether corresponding energy hierarchical information, determine in the targeted packets stream comprising echo bag.

Specifically, according to Figure 10, Figure 11 and formula 7 to formula 12, LSP is based on the invention provides one kind （Line Spectral Pair, line spectrum pair）The method for carrying out the detection of echoes of packet voice；If the echo determining unit 31 ' According to the reference packet stream of the correspondence call ends in correspondence reference buffer area, with reference to the targeted packets stream and the reference The direction of transfer information corresponding to each packet data package in stream of packets, using this method, determine the targeted packets stream with There is similar packet, then the echo determining unit 31 ' can be further combined with the targeted packets stream in reference packet stream The energy hierarchical information corresponding with multiple consecutive packet data packages corresponding in the reference packet stream（I.e. all kinds of increasings Benefit（gain）Information）, judge the similar packet in the targeted packets stream with the presence or absence of decay, i.e. energy hierarchical information Less than the energy hierarchical information of the corresponding reference packet stream；If in the presence of, prove that the similar packet wraps for echo, Then wrapped in the targeted packets stream comprising echo.

In step s32 ', echo cancellation devices are worked as comprising echo bag in the targeted packets stream, to the targeted packets Stream carries out echo cancellor, to obtain the elimination echo stream of packets corresponding with the targeted packets stream.

Specifically, if being wrapped in the targeted packets stream comprising echo, in step s32 ', echo cancellation devices are then to described Targeted packets stream carries out echo cancellor, for example, be replaced using bag is replaced to detected echo bag, with obtain with it is described The corresponding echo stream of packets of elimination of targeted packets stream.Wherein, the bag of replacing includes but is not limited to noise bag（For example, bag The packet of noise containing certain type, such as white noise, comfort noise）, noiseless bag（For example, space division group）, in targeted packets / 8th rate packets finally cached in stream etc., and its mixing.

It is obvious to a person skilled in the art that the invention is not restricted to the details of above-mentioned one exemplary embodiment, Er Qie In the case of without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter From the point of view of which point, embodiment all should be regarded as exemplary, and be nonrestrictive, the scope of the present invention is by appended power Profit is required rather than described above is limited, it is intended that all in the implication and scope of the equivalency of claim by falling Change is included in the present invention.Any reference in claim should not be considered as to the claim involved by limitation.This Outside, it is clear that the word of " comprising " one is not excluded for other units or step, and odd number is not excluded for plural number.That is stated in device claim is multiple Unit or device can also be realized by a unit or device by software or hardware.The first, the second grade word is used for table Show title, and be not offered as any specific order.

Claims

1. a kind of method for being used to be grouped acoustic echo elimination, wherein, this method comprises the following steps：

A obtains the source stream of voice packets for the call ends that pending packet acoustic echo is eliminated, wherein, the source stream of voice packets Include one or more packets packet；

B updates the targeted packets stream of the correspondence call ends in destination buffer according to the source stream of voice packets, wherein, The targeted packets stream includes the direction of transfer information corresponding to each packet data package in the targeted packets stream；

C is according to the reference packet stream for corresponding to the correspondence call ends in reference buffer area, with reference to the targeted packets stream and institute The direction of transfer information corresponding to each packet data package in reference packet stream is stated, carrying out echo to the targeted packets stream disappears Remove, to obtain the elimination echo stream of packets corresponding with the targeted packets stream；

D according to it is described eliminated echo stream of packets corresponding to direction of transfer information, the echo stream of packets that eliminated is sent Corresponding end into the call ends.

2. according to the method described in claim 1, wherein, the step b includes any one of following：

- according to the source stream of voice packets, determine the transmission corresponding to each packet data package in the source stream of voice packets Directional information；According to the source stream of voice packets, believe with reference to the direction of transfer of the packet data package in the source stream of voice packets Breath, updates the targeted packets stream of the correspondence call ends in destination buffer；

- according to the source stream of voice packets, update the targeted packets stream of the correspondence call ends in destination buffer；According to institute Targeted packets stream is stated, the direction of transfer information corresponding to each packet data package in the targeted packets stream is determined.

3. method according to claim 1 or 2, wherein, this method also includes：

- echo stream of packets has been eliminated according to, update the reference packet stream in the reference buffer area.

4. according to the method described in claim 1, wherein, the step b includes：

- the source stream of voice packets is respectively sent to the destination buffer and reference buffer area, buffered with updating the mark The targeted packets stream of the correspondence call ends in area, and the reference of the correspondence call ends divides in the reference buffer area Group stream, wherein, the targeted packets stream includes the direction of transfer corresponding to each packet data package in the targeted packets stream Information, the reference packet stream includes the direction of transfer information corresponding to each packet data package in the reference packet stream.

5. according to the method described in claim 1, wherein, the step c includes：

C1 is according to the reference packet stream for corresponding to the correspondence call ends in reference buffer area, with reference to the targeted packets stream and institute The direction of transfer information corresponding to each packet data package in reference packet stream is stated, determines whether wrapped in the targeted packets stream Containing echo bag；

C2 work as the targeted packets stream in comprising echo bag, to the targeted packets stream carry out echo cancellor, with obtain with it is described The corresponding echo stream of packets of elimination of targeted packets stream.

6. method according to claim 5, wherein, the step c1 includes：

- according to the reference packet stream of the correspondence call ends in correspondence reference buffer area, with reference to the targeted packets stream and institute State the direction of transfer information corresponding to each packet data package in reference packet stream, and with the targeted packets stream with it is described The corresponding energy hierarchical information of corresponding multiple consecutive packet data packages in reference packet stream, determines the target point Whether echo bag is included in group stream.

7. the method according to claim 5 or 6, wherein, the step c2 includes：

- when being wrapped in the targeted packets stream comprising echo, using replacement data bag, echo is carried out to the targeted packets stream and disappeared Remove, to obtain the elimination echo stream of packets corresponding with the targeted packets stream.

8. a kind of echo cancellation devices for being used to be grouped acoustic echo elimination, wherein, the equipment includes：

Acquisition device, the source stream of voice packets for obtaining the call ends that pending packet acoustic echo is eliminated, wherein, it is described Source stream of voice packets includes one or more packets packet；

Target update device, for according to the source stream of voice packets, updating the correspondence call ends in destination buffer Targeted packets stream, wherein, the targeted packets stream includes the biography corresponding to each packet data package in the targeted packets stream Send directional information；

Cancellation element, for the reference packet stream according to the correspondence call ends in correspondence reference buffer area, with reference to the mesh Stream of packets and the direction of transfer information corresponding to each packet data package in the reference packet stream are marked, to the targeted packets Stream carries out echo cancellor, to obtain the elimination echo stream of packets corresponding with the targeted packets stream；

Dispensing device, for having eliminated the direction of transfer information corresponding to echo stream of packets according to, is eliminated back by described Sound stream of packets sends the corresponding end into the call ends.

9. echo cancellation devices according to claim 8, wherein, the target update device is used for any one of following：

10. echo cancellation devices according to claim 8 or claim 9, wherein, the equipment also includes：

With reference to updating device, for having eliminated echo stream of packets according to, the reference packet in the reference buffer area is updated Stream.

11. echo cancellation devices according to claim 8, wherein, the target update device is used for：

12. echo cancellation devices according to claim 8, wherein, the cancellation element includes：

Echo determining unit, for the reference packet stream according to the correspondence call ends in correspondence reference buffer area, with reference to institute Targeted packets stream and the direction of transfer information corresponding to each packet data package in the reference packet stream are stated, the mesh is determined Whether mark in stream of packets comprising echo bag；

Echo cancellation unit, for when being wrapped in the targeted packets stream comprising echo, carrying out echo to the targeted packets stream and disappearing Remove, to obtain the elimination echo stream of packets corresponding with the targeted packets stream.

13. echo cancellation devices according to claim 12, wherein, the echo determining unit is used for：

14. the echo cancellation devices according to claim 12 or 13, wherein, the echo cancellation unit is used for：